Oracles in NP(NP) are Sufficient for Exact Learning

Johannes Köbler and Wolfgang Lindner

Abstract: We consider representation classes of polynomial query complexity in Angluin's exact learning model with all three possible combinations of membership and (proper) equivalence queries allowed. We show that in all three cases, polynomial query complexity implies already polynomial-time learnability where the learner additionally has access to an oracle in NP(NP). As an application, it follows that boolean circuits are polynomial-time learnable with equivalence queries and the help of an oracle in NP(NP). Our results are based on known combinatorial properties characterizing polynomial query complexity and a careful application of hashing-techniques.